architecture

Drafts and records v0 architecture decisions as durable ADRs with human ratification.

Updated Jul 10, 2026
One-click install
npx skills add https://github.com/subvisual/the-a-team --skill architecture-subvisual
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: architecture
Source: https://github.com/subvisual/the-a-team/tree/main/.claude/skills/architecture
Command: npx skills add https://github.com/subvisual/the-a-team --skill architecture-subvisual

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Before a v0 plan can mean anything, someone must decide the product's shape — repo structure, tech stack per surface, where the v0 runs, and the data strategy — and record those decisions durably instead of letting them dissolve into chat history. ## Core Features & Use Cases - Four-decision workflow: Walks repo shape, tech stack, v0 runtime location (local-first), and v0 data strategy in order, writing each answer as an ADR at docs/product/adr/NN-<slug>.md. - Precedence-based decision rule: Applies project binding > team default > ask, naming any lower rule that would have chosen differently in Alternatives considered. - Human ratification gate: Decisions that cost money, foreclose options, or contradict the target repo are stamped decided_by: human only after explicit approval; unanswered ones land as status: parked. - Use Case: After a dev review of a target repo, run this Skill to produce ratified ADRs covering monorepo shape, per-surface stack, local-first deployment, and mocked API-shaped data, each with alternatives, consequences, and a revisit-when trigger. ## Quick Start Use the architecture skill to decide and record the v0 repo shape, tech stack, runtime location, and data strategy as ADRs based on the dev review findings.

Frequently Asked Questions about architecture

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I record architecture decisions as ADRs for a v0 product?▼

Walk the four decisions in order — repo shape, tech stack, where the v0 runs, and v0 data strategy — and write each as one file at docs/product/adr/NN-<slug>.md using the ADR template. Each ADR needs context, a decision, at least one real alternative, consequences, and a revisit-when trigger.

What decisions must be made before planning a v0?▼

Four decisions: repo shape (one repo or monorepo, which surfaces exist), tech stack per surface, where the v0 runs (local-first by default), and the v0 data strategy (real integration versus mocked API-shaped data). Each constrains the next, so they are answered in that order.

Which architecture decisions require human approval?▼

Human ratification is required when a decision costs money, commits to a paid tier or vendor, forecloses an expensive-to-reopen option, contradicts a project binding, rests on a finding rated expensive or unknown, or picks a data strategy for a job whose value is live data.

How does precedence work between project bindings and team defaults?▼

Precedence is project binding > team default > ask. A fact in the target's A-Team Config or Technical context binds first; a Declared default applies openly with a confidence-stamped assumption; otherwise the question goes to the human with a recommendation.

When should I not use an architecture decision skill?▼

Do not use it for component or schema design, API contracts, library-level picks inside an already-settled stack, or implementation work. Those belong to the specification and development phases; this Skill only decides the v0's overall shape.

What happens if no human is available to ratify a decision?▼

The decision is not written as active. Its question goes to context.md's Awaiting answers section, the ADR lands with status: parked holding the draft and alternatives, and an open question is logged in research-plan.md while other decisions proceed.